Genome-wide analysis of mRNA stability using transcription inhibitors and microarrays revea

Copyright©2004,Americanϩ0DOI:Society10.1128/MCB.24.12.5534–5547.2004

forMicrobiology.AllRightsReserved.

Genome-WideInhibitorsControlandAnalysisMicroarraysofmRNAStabilityUsingTranscription

ofRibosomeRevealsBiogenesisPosttranscriptional

Factors

Jo¨rgGrigull,SanieMnaimneh,andTimothyJeffreyR.Hughes*

Pootoolal,MarkD.Robinson,

BantingandBestDepartmentToronto,OntarioofMedicalM5GResearch,1L6,Canada

UniversityofToronto,

Received2December2003/Returnedformodification6January2004/Accepted9March2004

transcriptionUsingDNAtested,tomicroarrays,rpb1-1wecomparedglobaltranscriptstabilityprofilesfollowingchemicalinhibitionofmicroarraytheeffectsofthiolutin(atemperature-sensitiveand1,10-phenanthrolinealleleofwereyeastmostRNAsimilarpolymerasetorpb1II).-1.Amongthefiveinhibitorsscriptionalincludesresponsedataalreadytostressesintheliteraturesuchasheatrevealedshock,similaritybetweenmRNAstabilityAcomparisonprofilesandtothevarioustran-andwereribosomeatransientcomponentsamongtheassembly,shutoffofgeneralmRNAtranscription.consistentGeneswithencodingthefactfactorsthattheinvolvedgeneralinstressrRNAsynthesisresponseleaststablewhichtranscripts.areoftenobservedWeexaminedtobecoordinatelytheeffectsofdown-regulateddeletionsofgenesinyeastencodingmicroarraydata,patternshowedofmRNACcr4pstabilityandPan2pafterandinhibitionputativeoftranscriptionRNA-bindingbyproteinschemicalsPub1pand/orandheatPuf4ponthegenome-widedeadenylasebothtranscriptionalribosomalthatCcr4p,proteinsthemajorandrRNAyeastmRNAsynthesisdeadenylase,andribosomecontributesassemblytothedegradationstress.oftranscriptsThisexaminationencodingmRNAsproteins.followingresponsetranscriptionaltoheatstress.Pan2pandPuf4palsocontributedfactorstoandthemediatesdegradationalargeratepartofofthesethemRNAstability.

Ourresultsindicatethatshutoff,theabundancewhilePub1pofribosomepreferentiallybiogenesisstabilizedfactorstranscriptsiscontrolledencodingattheribosomallevelofmRNAturnoverisanimportantfactorintheregulationofidentificationofcis-andtrans-actingfactorsaffectingmRNAgeneexpressionineukaryoticcellsandcomplementstranscrip-tionalregulationbyendowingthecellwiththecapabilitytostabilityhasledtothegeneralhypothesisthatregulatorsofrapidlyvarythelevelsofexistingtranscripts(7,27).mRNAmRNAstabilitymayrepresentafunctionalequivalentofbac-half-livesrangefrom3mintomorethan100mininSaccha-terialoperonsandeukaryoticpathway-specifictranscriptionromycescerevisiae(20,58)andfromϳ15mintomorethan10hfactors(30,31).Severalofthesetrans-actingfactorsconstituteinmammals(27).ThedetailsoftheeukaryoticmRNAdecayalinkbetweenmRNAstabilityandtranslation(27).Forex-pathwayarebestunderstoodinS.cerevisiae,wheremanyoftheample,atlowironconcentrations,theironregulatoryproteinmajorproteinsinvolvedhavebeenidentifiedandcharacterizedIRP-1(identicaltoaconitase)bindstotheiron-responsive(53,60).Deadenylationanddecappingarethetwomostim-element(IRE)inthe5ЈUTRofferritinmRNAandblocksportantstepsinthemRNAdegradationpathwayandtypicallytranslation,whileIREsinthe3ЈUTRofthetransferrinrecep-occursequentially(10,55,60),althoughdecappingisnotab-torcontroltranscriptstability(51).Similarly,AU-richele-solutelydependentupondeadenylation(54).Deadenylationisments(AREs)inthe3ЈUTRsofmanymammalianprotoon-carriedoutbythePan2/Pan3andtheCcr4/Caf1poly(A)nu-cogeneandchemokinetranscriptsmodulatetranscriptstabilitycleasecomplexes(54),anddecappingiscarriedoutbythe(12,13),andthebindingaffinityofproteinsinthemammaliandecappingenzymesDcp1and/orDcp2,allofwhicharecon-ELAVfamilytoAREscorrelateswiththestabilizationorservedamongeukaryotes(33,44).Adeadenylatedandde-destabilizationofthesemRNAs(9).However,cis-andtrans-cappedmRNAisdegradedbythe5Ј33ЈexonucleaseRat1actingdeterminantsofrelativestabilityhavenotbeenidenti-and/orXrn1(22)andbya3Ј35Јexonucleolyticproteincom-fiedformostmRNAsdespitethefactthattranscriptshaveplexknownastheexosome(26).

characteristicandwidelyvaryinghalf-livesthatoftencorrelateTranscriptstabilityistraditionallythoughttoberegulatedbywiththefunctionoftheencodedprotein(58).Forexample,inspecificsequencesorstructuresinthe3Јor5ЈuntranslatedcontrasttotextbookexamplesofmRNAstabilitydeterminantsregion(UTR)andbycognatetrans-actingfactorsthatrecog-suchasIREsandAREs,ithasbeenreportedthatthecisnize,andinsomecasesmaybindto,theseelements(40).The

determinantsresponsiblefortherelativelyshorthalf-livesofyeastribosomalproteintranscriptsmayresideinthecodingsequence(20).Coupledwiththefactthatcisdeterminantsofment*CorrespondingmRNAstabilitymaybesecondarystructuresorcompositeM5Gofauthor.Mailingaddress:BantingandBestDepart-sequencefeatures(ratherthanshort,contiguousmotifs),thismail:t.hughes@utoronto.ca.

1L6,MedicalCanada.Research,Phone:University(416)946-8260.ofToronto,Fax:(416)Toronto,978-8528.OntarioE-findingsuggeststhatsuchciselementsmaybedifficulttoiden-

5534

Downloaded from mcb.asm.org at Penn State Univ on February 5, 2008 VOL.24,2004tifyanddissectbyeithercomputationalapproachesorconven-tionaltechniquesinmolecularbiology.

GenomesequencinghasidentifieddozensofproteinswithRNA-interactingmotifs(e.g.,helicases,nucleases,andRNA-bindingproteins),andthesepotentialtrans-actingfactorsrep-resentanalternativestartingpointfortheidentificationofdeterminantsandmodifiersofmRNAstability.TothoroughlytesttheeffectofanyexperimentalperturbationonmRNAstabilityintrans,awhole-genomeapproachisneededthatcandetectoverallchangesinmRNAhalf-livesaswellasidentifywhetherspecificfunctionalclassesoftranscriptsareaffected.Twogeneralstrategieshavebeendescribedforusingmicroar-raystostudymRNAstability.Thefirststrategyistomeasuretherelativeabundanceoftranscriptsbetweentwoconditions,e.g.,awild-typestrainandamutantstrain(37).Steady-statemeasurements,however,failtocapturetheactualdynamicsofmRNAdecayandwillreflectbothprimaryandsecondaryeffects.Thesecondstrategyistomeasuretheabsolutehalf-livesofmRNAsoveratimecoursefollowingtheinhibitionoftranscription(8,18,45,58).Whilethisstrategyhastheadvan-tagethatdecayratesaremeasureddirectly,itrequiresasep-aratetimecourseandaseparatesetofarraysforeachsampleanalyzed,anditdoesnottakeadvantageofthestrengthofthetwo-color(i.e.,redandgreen)microarraysystemindetectingsmallchangesinrelativetranscriptabundance.

AtroublesomecaveatofanymRNAdecayexperimentthatinvolveseithergeneticorchemicalinhibitionoftranscriptionisthatsideeffectsofthesetreatments,andpossiblytheactofshuttingofftranscriptionitself,mayaffectmRNAdecay.Forexample,manystudiesofmRNAstabilityinyeast(20,38,41,58)haveemployedtherpb1-1allele(42),atemperature-sen-sitivemutantinthecatalyticsubunitofRNApolymeraseII(PolII)(2).Shutoffwithrpb1-1requiresatemperatureshift,whicheveninawild-typestraincausesarapiddecreaseinmRNAlevels(21,38).ThisdecreaseappearstobeduetoarapidandtransientshutoffofgeneraltranscriptionanddecayofmRNAsatthenaturalrate(39).Levelsofribosomalproteintranscriptsareaffectedmostprominently(21,32,38,58);inadditiontohavingshortinherenthalf-lives,thesetranscriptscompriseroughly50%ofthemRNAofactivelygrowingyeast(38).Thetranscriptioninhibitorsthiolutinand1,10-phenanth-roline,likeheatandmanyotherenvironmentalperturbations,alsoelicitthestressresponsetranscriptionalprogramatleastpartially(1,11,15).

Theexperimentswedescribeherewereinitiatedwiththepurposeofestablishingageneralmicroarray-basedmethodforanalyzingtheeffectsofknownorpotentialtrans-actingfactorsonmRNAdegradationand/orstabilizationinwhichrelativechangesinabundancebetweenthemutantandwildtypearecomparedafterinhibitingtranscriptioninbothculturesinor-dertoidentifysubsetsoftranscriptsthatarepreferentiallyaffected.WebeganbyexaminingthespecificityofapaneloftranscriptioninhibitorsusingDNAmicroarraystocomparetheireffectstothoseofrpb1-1.Weobservedthatthepoly(A)ϩmRNAdecayprofilebearsastrongresemblancetotheheatshockresponse,consistentwithamajorroleformRNAdecayindeterminingtheheatshockresponse.Wealsoobservedthattranscriptsencodingribosomebiogenesisfactorsareamongtheleaststableinthetranscription-inhibitedyeastcell.Wethenexemplifiedthesystemfortheanalysisoftrans-acting

TRANSCRIPTIONINHIBITORSANDMICROARRAYS5535

factorsbycomparingpub1-⌬,ccr4-⌬,puf4-⌬,andpan2-⌬strainstothewildtypebeforeandaftertreatmentwithtran-scriptioninhibitors(andheatshockinthecaseofccr4-⌬).ThisanalysisconfirmedthatCcr4pisamajormediatorofpoly(A)ϩmRNAstabilityinyeastandalsoshowedthatCcr4patleastpartiallymediatestheheatshockresponse.Thestabilityofribosomebiogenesisfactorsorribosomalprotein-encodingtranscriptswasaffectedbyallfourofthemutantsanalyzed.Thisresultshowsthatthedown-regulationofmRNAsencod-ingribosomebiogenesisfactorsfrequentlyobservedinmi-croarraydataaredueatleastinparttotheircoordinatereg-ulationattheleveloftranscriptstabilityandthatthisrepresentstheposttranscriptionalfunctionalequivalentofabacterialoperonasproposedbyKeeneandTenenbaum(30,31).

MATERIALSANDMETHODS

Strains,yeastculture,RNAextraction,andlabeling.Therpb1-1strainYF2475(MATarpb1⌬187::HIS3[pRP1-10;rpb1-1onaLEU2/CEN/ARSplasmid])wasagiftofDaveJansma(UniversityofToronto).AllotherstrainswerehomozygousdiploidsfromtheSaccharomycesGenomeDeletionscollection;thewild-typestrainusedinallexperimentswasBY4743,whichisthewild-typediploidforthedeletionsconsortiumstrains(16).Culturesweregrowntoafinaldensityofϳ107cells/mlinyeast-peptone-dextrosemediumat30°Cwithvigorousshaking,withtheexceptionoftherpb1-1strain,whichwasgrowninyeast-peptone-dextrosemediumat25°C(permissivetemperature);transcriptionwasshutoffbytrans-ferringtheculturefromthepermissive(25°C)tothenonpermissive(37°C)temperaturebyaddinganequivalentvolumeofmediumwarmedto49°C.Attheindicatedtimepoints,cellswereharvestedby2minofcentrifugationat3,000rpminatabletopcentrifuge(Eppendorf5810)atroomtemperatureandimme-diatelyfrozeninliquidNprecipitation,andmRNA2.RNAwasextractedbyhotphenolfollowedbyethanolwaspurifiedonoligo(dT)cellulose(NewEnglandBiolabs).Twomicrogramsofpoly(A)ϩmRNAwasreversetranscribedforcDNAsynthesisandlabeledwithCy3andCy5dyesasdescribedpreviously(24).Chemicalinhibitors.AlldrugswerepurchasedfromSigma-Aldrichexceptforthiolutin,whichwasagiftfromPfizer.Thefollowingdrugswereused:1,10-phenanthroline(100␮g/mlinethanol),6-azauracil(2mg/ml),thiolutin(3␮g/mlindimethylsulfoxide),ethidiumbromide(200␮g/ml),andcordycepin(33␮g/ml).

Microarraymanufactureanduse.TheOperonYeastOligoset,whichincludesa70-meroligonucleotiderepresentingeachof6,300yeastopenreadingframes,wasdilutedtoafinalconcentrationof1mg/mlinasolutionof50%dimethylsulfoxideand0.05%sodiumdodecylsulfateandspottedontopoly-L-lysineslidesbyusingaroboticspotterwith16pins(Virtek,Toronto,Canada).Blocking,hybridization,andwashingproceduresweredoneasdescribedpreviously(19),withtheexceptionsthatwasheswererestrictedto20seach.Allmicroarrayhybridizationswereperformedinduplicate,withfluorsreversedonthesecondarray.SlideswerescannedonanAxonGenePix4000scanner.

Dataanalysis.Allmicroarrayfeatureextraction,normalization,clustering,statisticalanalyses,graphs,andfiguresweregeneratedwithMatlab(Mathworks).Initialspotintensitiesandratiosforeacholigonucleotideweredeterminedbythemedianofpixelswithineachspotafterlocalbackgroundsubtraction.Normal-izationfollowed,accordingtothemethodofYangetal.(62),wherebyalowesssmootherisappliedtotheratiosofeachexperimentoverintensity.

Geneontology(GO)annotationsweredownloadedfromftp://genome-ftp.stanford.edu/pub/yeast/(accessedNovember20,2002);eachannotationwas“propagatedupward”alongallparentalbranchesoftheGOgraph.

TheKolmogorov-Smirnov(KS)goodness-of-fitstatisticwasusedtodetect,foranyfunctionalcategoryofsizen,deviationsofthemRNAintensityratiorankdistributionsfromtheuniformdistributionandisgivenbytheequationDϭmaxԽFOϪFEԽ,whereFOandFEdenotetheobservedandexpectedcumulativedistributionfunctions,respectively.TheKStestisdistributionfreeandapplica-bleforsamplesizesofatleast35(47).ForanyGOcategory,thecalculatedPvaluerepresentsthesignificancelevelsoftheKStestatwhichthenullhypothesisofuniformrankdistributioncanberejected.

InthecalculationofPvaluesinFig.2and3,weusedtheasymptoticallynormaldistributionoftheSpearmanrankcorrelationcoefficientinthetestforindependence(17).

Half-liveswerecalculatedfor3,966mRNAswiththehighestspotintensitiesat

Downloaded from mcb.asm.org at Penn State Univ on February 5, 2008 5536GRIGULLETAL.MOL.CELL.BIOL.

tϭ0(medianofϾ50counts/pixelafterbackgroundsubtraction,wherecountsscalefrom0to65,535).Forthetimepoints5,12.5,and30mininthetworpb1-1establishedthatdeadenylationisthekeystepthatdeterminesexperiments,theintensitieswerenormalizedtothesumofallspotintensitiesandmRNAdecayandisusuallythestepthatisregulated(forreweightedwithafactor2Ϫt/␶_total,wherethehalf-life␶_reviews,seereferences10and55).Consistentwiththis,Wangassumedtobe23min.RatiosbetweentheseintensitiesandtotaloftotalmRNAwastheintensitiesattϭetal.(58)showedthattherpb1-1decayprofileisverysimilar,0werecalculatedandaveragedoverrpb1-1trials1and2.ForeachmRNAspeciesi,itsaverageratiosrwhetherthelabelingreactionwasprimedwitholigo(dT)ori(t)fortϭ5,12.5,and30minwereequatedwith2Ϫt/␶_i(t),where␶_randomprimers.

thelifetimeformRNAi(t)denotesthehalf-lifeofmRNAiinferredfromtimepointt;speciesiwasdeterminedastheaverageoverthethreeAswithmosttwo-colormicroarrayexperiments,thedatapointwiseinferences,␶_i(t).

resultingfromeachofthesehybridizationsarerelativeratiosDataavailability.Thefollowingdatacanbedownloadedfromhttp:foreachgenerelativetoitsabundanceinthestartingsample//hugheslab.med.utoronto.ca/Grigull/:allprocesseddatafilesfromallexperi-ments,anassembledratiofileforallexperiments,datatablesunderlyingeachofandalsorelativetotheaveragetranscript.Thesedataarethefigures,atableofPvaluesfromtheKStestforall342functionalcategoriesexpressedonalogscalewherepositivenumbersindicateanovertheranksingenessharedbyFig.1andtheWangetal.studydata(58),andincreaseinrelativeabundanceandnegativenumbersindicatetheSpearmanPvaluesamongallofourexperimentsandthoseoftheGaschetarelativedecrease.Figure1Bshowsthedatafor5,245genesal.(15)andWuetal.(61)studies.

rankedbytheiraverageratio(whichshouldrepresenttheirRESULTS

relativestability)overmultipleexperiments.Asanexample,inourfirstrpb1-1timecourse,thelogratiosatthe30-mintimeAnalysisofrpb1-1.Webeganbyanalyzinganrpb1-1strainaspoint(whichisclosetotheestimatedaverage23-minhalf-lifeareferenceforthechemicalinhibitorsoftranscription.SinceofyeastmRNAs[58]),theHSP12transcripthasalogratiooftherpb1-1strainisslowgrowing,evenatthepermissivetem-1.40(101.4perature,itislikelythatithasalterationsinsteady-statetran-script),indicatingϭ25-foldthatincreaseditisarelativelyrelativetostabletheaveragetranscript.tran-Inscriptlevelsincomparisontothewildtype,asistypicalforcontrast,RPA135(asubunitofPolI)hasalogratioofϪ1.22mutantswithagrowthdefect(24).Therefore,weusedthe(16-foldreducedrelativetotheaveragetranscript),indicatingprotocolillustratedinFig.1A,inwhichmRNAfromtimethatitisaveryunstabletranscript.Allofthemicroarraydatapointsfollowingthetemperatureshift(orapplicationofthedescribedhereinareavailableintheonlinesupplementalma-inhibitorforexperimentsbelow)wereeachcomparedtotheterialathttp://hugheslab.med.utoronto.ca/Grigull/,includingstartingculturewithDNAmicroarrays.Inthisexperimentalratios,spotintensities,andestimatesofmeasurementerror.design,2␮gofpoly(A)ϩmRNAisusedateachtimepointtoAlthoughitwasnottheprimarypurposeofourstudy,itiscreatecDNA.However,becausethepoolofmRNAisdead-possibletocalculatethehalf-livesofindividualpoly(A)ϩenylatingoverthetimepoint,theactualpoolofmRNAinthemRNAsfromourdatabyassumingamassdecayrate(inthiscellthatisselectedonoligo(dT)cellulosewilldecreaseandwillcase,tbeenrichedinmRNAswithslowdeadenylationratesanddatamirrored1/2ϭ23min).thedistributionTheoverallindistributiontheWangofethalf-livesal.studyin(58)ourongoinglow-leveltranscription.Consequently,therelative(Fig.2A),withmosttranscriptshavinghalf-livesbetween10ranksofratiosshouldreflecttherelativeranksofdeadenyla-and20minbutsomehavingmuchlongerhalf-lives.Disagree-tionrates(i.e.,stability)becausethemRNAsthataredead-mentsbetweenthetwostudiesmaystemfrommeasurementenylatedmoreslowlythanaveragewillbemorestableand,error;thereisathreefoldhighercorrelationbetweenthetwohence,theirproportionwillgraduallyincreaserelativetothatstudieswithgenesthatarehighlyexpressedinourdata(RϭoftheaveragemRNA,whereasthosethatarerapidlydead-0.48atthe90thpercentileofmicroarrayspotintensity)com-enylatedwillbelessstableandwillgraduallydecreaseinpro-paredtothosethatareweaklyexpressed(Rϭ0.15atthe10thportiontotheaverage.Thisdifferssubstantiallyfromthepro-percentile)(Fig.S1).Thetwostudieswereingoodagreement,tocolusedbyWangetal.(58),inwhichatwo-colormicroarrayhowever,whenhalf-liveswereaveragedforallgeneswithinanysystemandanrpb1-1strainwerealsoemployedbutwithgivenfunctionalcategory(Fig.2B).Thetwostudieshadsim-genomicDNAasthereferencesampleonthemicroarrays,ilarspreadsinhalf-liveswithinindividualcategories(Fig.S2).ratherthanmRNAfromthestartingculture.

AssumingthatpoormeasurementstendtoincreasethespreadToincreasesignalintensityonmicroarrayspotsandtore-withinacategory(i.e.,torandomizethemeasuredhalf-lives),ducepotentialcross-hybridizationfromnoncodingRNAsthisindicatesthatneitherstudyismoreprecisethantheother(whichmakeupthevastmajorityofcellularRNA),weusedoverall.

equalamountsofoligo(dT)-primed,poly(A)ϩ-selectedmRNAAmoregeneralwayofcomparingmicroarraydatafromineachchannelofeacharray.Thus,ouranalysiswasrestricteddifferentsourcesistouserelativeranks.TheSpearmanranktomRNAsthatcontainpoly(A)tails;subsequenteventsfol-correlationcoefficientmeasuresthestrengthoftheassocia-lowingdeadenylationwerenotobserved.Previousstudieshave

tionsbetweentwovariables(36)andisconceptuallysimilarto

cataloguedFIG.1.Comparisonofchemicalinhibitorsoftranscriptiontorpb1-1acoursetwo-colorbyrpb1experimentsassayGOcomparing(4,25).(A)followingsuccessiveSchematictheproceduretimediagrampointsofdescribedtoourandrelationshipoftranscriptstabilitiestoclassesofgenefunctionstheexperimentalstrategytocomparegeneticandchemicalinhibitionoftranscriptionusingforfipanelrsttimeA.pointThegenes(tϭ0).were(B)orderedLeftpanel,accordinglogratiototheof5,245mediantranscriptsrankofineachninegenedifferentamongtimemeasures-1,1,10-phenanthroline,datawereomitted.Centerandpanel,6-azauracilratiosoverexperiments.aheatshockApproximatelytimecourse1,000(15)areyeastshown,geneswithnotthemeetinggeneorderminimummaintainedintensityandspotqualitythestability,atleft.withRightgenepanel,orderdensityagainmaintainedisshownforfromeighttheofthemicroarrayGOannotationdataatclassestheleft.

identifiedashavingnonrandomdistributionsfromofrelativethemicroarraytranscriptDownloaded from mcb.asm.org at Penn State Univ on February 5, 2008 TRANSCRIPTIONINHIBITORSANDMICROARRAYS

5537

Downloaded from mcb.asm.org at Penn State Univ on February 5, 2008 VOL.24,2004

5538GRIGULLETAL.MOL.CELL.BIOL.

comparedFIG.2.toStatisticalthecalculatedanalysishalf-livesofmRNAreportedstabilitybyandWangcomparisonetal.(58).toThedatagenesinthewereliterature.selected(A)onCalculatedthebasisthathalf-lives(i)theirforspot2,867intensitiesgenesinatourtϭdata0in

Downloaded from mcb.asm.org at Penn State Univ on February 5, 2008 VOL.24,2004TRANSCRIPTIONINHIBITORSANDMICROARRAYS5539

thePearsoncorrelation(whichisnormallyusedinmicroarrayinteractionbetweenRNApolymeraseanditsDNAtemplatedataanalysis);essentially,themeasurementsbeingcompared(49).Chemicalconcentrations(seeMaterialsandMethods)(e.g.,microarrayratiosortranscripthalf-lives)areplacedonwerechosenasthefirstinaconcentrationseriesthatpre-anequivalentscalebyconversiontorelativeranks(Fig.2C).ventedgrowthofovernightcultures(datanotshown)inorderImportantly,thelikelihood(i.e.,Pvalue)thatagivenSpear-tominimizesecondaryeffects.Toourknowledge,ithasnotmancorrelationwouldbeobtainedbychanceiseasilycalcu-beendemonstratedthatcordycepinandethidiumbromidearelated,whichisnotthecaseforthePearsoncorrelation.TheeffectiveinhibitorsofPolIItranscriptioninyeast,althoughSpearmancorrelationisalsomorerobustagainstoutliers(e.g.,theydoinhibitgrowth.Forthiolutin,1,10-phenanthroline,andsporadicerrorsinmicroarraydata).TheSpearmanrankcor-6-azauracil,however,weconfirmedbydotblotanalysisoftotalrelationbetweenourdataandtheWangdata(58)is0.35and,RNAwithanoligo(dT)probethatchemicaltreatmentresulteddespitenumerousoutliers(i.e.,pointsthatdisagreebetweeninareductioninmRNAlevelswhich,likethoseinrpb1-1(butthetwostudies;Fig.2C,upperleftandlowerrightcorners),unlikethoseinheat-shockedsamples),didnotrecoverafter80hasaPvalueofϽ1.0ϫ10Ϫ101.Asareference,for424min(datanotshown).

differentmicroarrayexperimentsintheliterature(Wuetal.Figure1Bshowsthemicroarrayratiosobtainedfor5,245[61]andreferencestherein),thePvaluesforthe,676pos-yeastgenesovereachtimepointinallnineexperimentssibleϪ101rankcorrelationsareshowninahistograminFig.2B;(rpb1-1,1,10-phenanthroline,and6-azauracilwereeachre-10wouldbewellwithinthetop5%.These424experi-peated,andthecordycepin5-mintimepointsamplewaslostmentscontaintimecoursesofsporulation,mating,andthecellduetotechnicalreasons).ThePvaluesoftheSpearmanrankcycleaswellasmultipleinstancesofdifferentmutantsinthecorrelationcoefficientsamongalloftheseexperimentsaresamepathway;hence,afewpercentofthecorrelationsaregiveninthesupplementalmaterials.Apositivecorrelationexpectedtobehighlysignificant.

withtherpb1-1datawasobtainedwithallofthechemicalThisanalysisalsorevealedotherexamplesinwhichrepli-inhibitorsoftranscriptionexceptethidiumbromide.Thehigh-catesofthesamemicroarrayexperimentbytwodifferentlabsestcorrelationstorpb1-1wereobtainedwith1,10-phenanth-yieldedhighlysignificantcorrelationsbutwithmanydisagree-rolineandthiolutinaswellas6-azauraciltoaslightlylessermentsregardingindividualgenes;forexample,Fig.2Dshowsextent(Fig.3andsupplementalmaterial).Inaddition,theseacomparisonbetweentheheatshockseriesofGaschetal.(15)chemicalscorrelatedmorecloselywitheachotherthanwithandthatofCaustonetal.(11),whichhaveacorrelationcoef-rpb1-1(Fig.3),suggestingthateither(i)thedrugssharesec-ficientofonly0.41.Insummary,theseanalysesconfirmedthatondaryeffectsor(ii)rpb1-1isitselfnotanidealmodeloftheexperimentalprotocolusedhereyieldsrelativehalf-livestranscriptionalshutoffinawild-typecell.

thatareconsistentwithpreviousobservations,particularlywithWealsoexaminedthecharacteristicdifferencesbetweentheregardtogroupsoffunctionallyrelatedgenes,andthefactthatrelativeranksoftranscriptstabilitiesobtainedwithrpb1-1andthereisnotabsoluteagreementoneverymeasurementistyp-differenttranscriptioninhibitorsinordertofurthercompareicalofcomparisonsamongmicroarraystudies.

them(Fig.4).Weidentified104geneswhoserankpositionsComparisonoffivechemicalinhibitorsoftranscriptiontowereskewedintherpb1-1,thiolutin,1,10-phenanthroline,andrpb1-1.Wenextexaminedtheeffectsoffiveknownorputative6-azauracilexperimentsandgroupedthembyhierarchicalchemicalinhibitorsofPolIItranscriptionwiththesameex-clustering.Wethenassessedtheresultingclustersforastatis-perimentalprotocol(Fig.1A).Thiolutinisatranscriptionin-ticallysignificantenrichmentofanyfunctionalcategoriesusinghibitorthatinteractswithallthreeRNApolymerases(52).thehypergeometricPvalue(46,50).Therpb1-1experiments1,10-Phenanthrolineisametalchelatorthatmostlikelyinhib-specificallyfeaturedrelativeup-regulationofTytranscriptsitsPolIIbysequesteringmagnesium(29).6-Azauracilisa(i.e.,transposons[Fig.4,lightblue]).ThisresultmaybeduetonucleotidebaseanalogthatstronglyaffectsintracellularGTPthetemperatureshift(which,amongourexperiments,waslevelsandiscommonlyusedtoscreenformutantsaffectinguniquetorpb1-1).However,inductionofTyelementswasnottranscriptelongation(14,48).Cordycepinisanadenosinean-seeninheatshock(15).Wealsoobservedthattherpb1-1strainalogthatisincorporatedintotranscriptsasifitwasanormalgrowsslowlyevenatthepermissivetemperature;hence,rpb1-1basebutlacksa3Јhydroxylgroupthatcanattackthealphamaybedysfunctionalatanytemperature.1,10-Phenanthrolinephosphorusontheincomingbase.ThisprocessleadstochaincausedapparentstabilizationofagroupoftranscriptshighlyterminationbecausetheRNApolymerasehasnoproofreadingenriched(PϽ0.000016)forthefunctionalcategory“ionicability(M.Crowley,personalcommunication).Ethidiumbro-homeostasis”(46)(Fig.4,highlightedinred).ThismaybemideisaDNAintercalaterthatpresumablyinterfereswithphysiologicallyrelevant(i.e.,thesetranscriptsmaybeeither

ourhalf-livesdataarebyinbothabovestudies50countsfor195perfunctionalpixelandcategories(ii)theyarefrompresenttheGOintheBiological4,686genesProcessfordatabasewhichWangetal.reportedhalf-lives.(B)MedianmRNAinranksour20ormoremembers.(C)ComparisonofmRNAstabilityranksfromourrpb1-1data(medianthatranksarerepresentedforthe5-,12.5-,intheandsetof30-min2,867timetranscriptspointsstable.aretwopossible(D)inrpb1Analogousincreasing-1experiments)comparisonorderofandstability;therankedofdatai.e.,fromgenesstabilitiesGaschinetthereportedal.lowerbyWangetal.(58)usingtheSpearmanrankcoefficientover3,736genes.The(15)leftandcornerCaustonareettheal.(11).most(E)stableHistogramandthoseofinSpearmantheupperrankrightPcorneraretheleastrepresentationcomparisonsproportionoftheamongKStest.424Proceedingdifferentfrommicroarraythelowestexperimentstothehighest(61)usingrank,thethesameproportion3,736genesasdescribedforpanelvaluesA.(F)amongSchematic,676numbersareofmoreallgenes.stable.

Deviationfromthediagonalindicatessignificance(seeMaterialsandMethods).ofgenesinAsaforgivenpanelcategoryA,genesiscomparedwithlowertoranktheDownloaded from mcb.asm.org at Penn State Univ on February 5, 2008 5540GRIGULLETAL.MOL.CELL.BIOL.

etFIG.Wangal.(15).3.Diagrametal.AdatatotalshowingPvaluesofSpearmanrankcorrelationsamongexperimentsinthisstudyandtheheatshocktimecourseofGasch(58),of3,736andthegenesGaschwereetusedal.datatodetermine(15).Phen.,the1,10-phenanthroline;correlations;theseare6-azaU,thegenes6-azauracil;presentandrep,consideredreplicate.

“good”forourdata,theDownloaded from mcb.asm.org at Penn State Univ on February 5, 2008 VOL.24,2004TRANSCRIPTIONINHIBITORSANDMICROARRAYS5541

selectedFIG.4.nomissingthatDifferencesdatahadpoints.anincreaseinrelativeThedifferencesordecreasetranscriptininstabilitiesamongrpb1-1,thiolutin,1,10-phenanthroline,and6-azauracil.Atotalof104transcriptswererelativerelativeranksrankofrelativeatleastto30the(relativemediantoaretheshownmedianinrank)theclustergram.

inatleast2ofthe28timepointsshown,withDownloaded from mcb.asm.org at Penn State Univ on February 5, 2008 5542GRIGULLETAL.stabilizedorup-regulatedbyanyresidualtranscriptionalactiv-ityinthetranscription-inhibitedcells)since1,10-phenanthro-lineisametalchelator.Thiolutinhadasimilareffect,andthisraisesthepossibilitythatthiolutinmayimpactmetalions,whichtoourknowledgehasnotbeenpreviouslyproposed.6-Azauracilaffectedagroupoftranscriptsenrichedforgenesinvolvedinpurinemetabolism(PϽ3ϫ10Ϫ7)(Fig.4,high-lightedindarkblue).ThesetranscriptsincludeIMD1,IMD2,IMD3,andIMD4,thefouryeastgenesthateachencodeanIMPdehydrogenase,thetargetof6-azauracil.TheinductionofIMD2inresponseto6-azauracilhasbeenpreviouslydescribedandisbelievedtoberegulatedattheleveloftranscriptionsinceitisdependentupontranscriptionelongationfactors(48).Hence,thetranscriptionalshutoffby6-azauracilisprob-ablyincomplete.

Relationshipbetweentranscriptionalshutoffandstressre-sponse.Toaskwhetherthepoly(A)ϩmRNAdecaysignatureisamajorcomponentofothermicroarrayexpressionprofilesintheliterature,andalsotoexaminetherelationshipbetweenourpoly(A)ϩmRNAdecayprofilesandpreviouslyreportedheatandstressresponsesintheliterature,wenextcomparedourpoly(A)ϩmRNAdecaydatatoover450othermicroarrayexpressionprofilesintheliterature(11,15,61)(Fig.S3anddatanotshown)usingthePvalueoftheSpearmanrankcor-relationcoefϩficientsasdescribedabove.Correlationsbetweenourpoly(A)mRNAdecayprofilesweremostsimilartore-sponsestoenvironmentalstressessuchasheatandnutrientdeprivation(Fig.3anddatanotshown).Thisresultisconsis-tentwithmRNAdecaybeingamajormediatoroftheheatshockresponse(39)andwithapreviousreportthat1,10-phenanthrolineatleastpartiallyinducesthestressresponse(1).Hence,themRNAdecayprofilefollowingtranscriptionalshutoffmaybeexperimentallyinseparablefromthestressre-sponse;likewise,whatislearnedbystudyingthemRNAdecayprofilemayalsoapplytothestressresponse.

Correlationsbetweengenefunctionandtranscriptstabilityfollowingtranscriptioninhibition.Anintriguingaspectofthecorrespondencebetweenpoly(A)ϩmRNAdecayprofilesandthoseofheatandotherstressesisthepossibilitythatstress-inducedtranscriptionalalterationsingroupsoffunctionallyrelatedtranscriptsmaybemediatedattheposttranscriptionallevel.Inparticular,Gaschetal.(15)notedthedown-regulationofmanyribosomebiogenesisfactors,ashaveothers(61);how-ever,themechanismcontrollingthisregulationhasnottoourknowledgebeenconclusivelyidentified.Wangetal.(58)notedmanycorrelationsamongthestabilitiesoftranscriptsthaten-codecomponentsofproteincomplexes,althoughribosomebiogenesisfactors,asidefromtheribosomalproteinsthem-selves,werenotspecificallymentioned.

Tosurveythegeneralcorrelationbetweengenefunctionandtranscriptabundanceindifferentstudies,weappliedtheKStest(47),whichestimatesthelikelihood(i.e.,assignsaPvalue)thatthedistributionofasubsetofranksisnonrandom(i.e.,thatthedistributiondeviatesfromtheexpecteduniformdis-tributionthatwouldbeobtainedbyassigningranksatran-dom).Figure2FillustratestheKStest.Theexpecteddistribu-tion,ifthegenesinagivenfunctionalcategoryappearatrandomamongtheranksofstabilities,isalinealongthediagonal;thisisshownfor“budding”whichisarandomlyselectedfunctionalcategorywithaPvalueofϳ1(i.e.,notMOL.CELL.BIOL.

significant)bythistest.Incontrast,mRNAsencodingproteinsinvolvedincellularrespirationandglucosemetabolism,asub-setofwhichhaspreviouslybeennotedtobecontrolledatthelevelofmRNAstability(34),areextremelysignificantinbothourdataandtheWangetal.data(Fig.2BandF),tendingtobemorestablethanwouldbeexpectedbychance.

Wetestedtherankdistributionsof304GOcategories(4,25)with35ormoregenesinthecategory(arule-of-thumbcutoffintendedtoavoidsamplingerror)(47).Amongthe304categories,63weresignificantatPϽ0.001inourdata(sup-plementalmaterials).Amongthemostsignificantcategoriesweobservedwereribosomebiogenesisfactorsandribosomeassemblyproteins;inbothourdataandtheWangetal.data(58),theseproteinsappeartobeamongtheleaststableyeasttranscripts(Fig.1Band2BandF).

Examinationofrelativetranscriptstabilitiesinccr4-⌬,pan2-⌬,pub1-⌬,andpuf4-⌬.Finally,weappliedtheprotocoloutlinedinFig.5Atocomparerelativetranscriptstabilitiesbetweenmutantandwild-typestrainsinordertoidentifypo-tentialtrans-actingfactorsthatspecificallyregulatedistinctfunctionalcategories.Thefourmutantstrainsweanalyzedwereselectedonthebasisthatthemutatedgenesencodeknownorpotentialregulatorsoftranscriptstability.Ccr4pandPan2parecomponentsofthetwoyeastmRNAdeadenylases(54).Pub1pistheyeasthomologofHuRintheELAVproteinfamily,whichbindsAREs(9,56).Puf4pisoneoffiveproteinsinyeastthatcarrythePumiliodomain,anRNA-bindingdo-mainthatmediatesbindingtospecificmRNAsinatleastoneotheryeastprotein(Puf3p),promotingdecay(43).Briefly,thewild-typeandmutantculturesweregrownsimultaneouslytosimilardensities;chemicalinhibitionwasinitiatedinbothcul-turesjustaftertimezero,andmRNAfromidenticaltimepointsinthetwocultureswascomparedinthetwochannelsbyusingmicroarrays.Thepub1-⌬andpuf4-⌬strainswereexam-inedonlyoncewith1,10-phenanthrolineasthetranscriptionalinhibitor.Theccr4-⌬andpan2-⌬strainswereeachexaminedtwice,firstbyusing1,10-phenanthrolineasthetranscriptionalinhibitorandthenbyusing6-azauracil.Thesetwochemicalswerechosenbecausetheyarecommerciallyavailableandtheirmechanismsofactionarewellcharacterizedbutdifferentfromeachother.ThemicroarraydataarerepresentedinFig.5Bandshow(i)thattheabsenceofanyoftheseproteinsfromthecellaffectsmRNAstabilityinspecificwaysand(ii)thatthisobser-vationisreproducibleovermorethanonetranscriptioninhib-itor;verysimilarresultswereobtainedwith1,10-phenanthro-lineand6-azauracil.Wealsoverifiedthatinccr4-⌬,essentiallythesameresultsareobtainedbyusingrandom-primedtotalRNA;i.e.,theeffectsweobservedforpoly(A)ϩmRNA,whicharepresumablyduetodifferencesindeadenylationratesbe-tweenthewildtypeandthemutant,arereflectedintheoveralldecay(Fig.5B).

TherearethreewaysthatratiosforspecificmRNAscanariseinthisprocedure.Thefirstisassteady-statealterationsintranscriptabundancethatarepresentbeforetranscriptionisinhibited;thesewillincludeprimaryandsecondaryeffectsaspreviouslynoted(37).Steady-statealterationscanbeidentifiedbecausetheyshouldbepresentinthefirsttimepointand,iftheyarenotaresultofalteredmRNAstability,theirratioswillnotchangesubstantiallyoverthetimecourse.Thesecondwayratioscanariseisiftheperturbationcausesanoveralleffecton

Downloaded from mcb.asm.org at Penn State Univ on February 5, 2008 afterFIG.withomitted.ratioschemical5.Comparisonoftwofoldinhibitionofthespecificeffectsofgeneticperturbations,knownorsuspectedtoaffecttranscriptstabilities,followingatimecourseorofmoretranscription.inthefour(A)orSchematicdiagramsoftheexperimentalmethod.(B)Leftpanel,clusteringanalysisof1,115genesThegeneRightorderpanel,(verticaldensityaxis)isisshownidenticalfor11fromGOmoreleftannotationtimepointstoright.classesamongPhen.,enrichedtheexperiments1,10-phenanthroline;inoneormoreshown.Genesthataretransposons(Tyelements)were6-azaU,clusters6-azauracil.

accordingtothehypergeometricPvalue(50).5543

Downloaded from mcb.asm.org at Penn State Univ on February 5, 2008 5544GRIGULLETAL.mRNAhalf-lives;thiswillhavetheeffectofagradualalter-ationinratiosthatisproportionaltomRNAhalf-life,becausetranscriptswithshorthalf-liveswillhaveagreaterproportional“gain”inamountatanygiventimepoint.Thesealterationscanbeidentifiedbecausetheslopesoftheratiosovertimeshouldbeproportionaltothehalf-livesofthemRNAsoveratleastthepartofthetimecoursewhereaccuratemeasurementscanbeobtained.ThethirdwayratioscanariseisbyanalterationinthedegradationratesofspecificmRNAclassesinthemu-tant.

Ccr4pexemplifieshowwedistinguishedbetweentheseef-fects.First,thezerotimepoint(i.e.,steadystate)containsmanytranscriptionalalterationswhichdonotchangeappre-ciablyoverthetimecourse;forexample,agroupoftranscriptsencodingpredominantlymitochondrialproteinsareindicatedbyanorangebarnearthetopofFig.5B.Second,visualin-spectionsuggeststhatintheccr4-⌬mutant,mRNAstabilityisaffectedalmostindependentlyofthegenericstabilityofthemosttranscriptsfollowingtranscriptionalinhibition(Fig.5B;comparetheheatshockcolumnswiththe1,10-phenanthrolineand6-azauracilcolumns).Thisideaisconfirmedbythestatis-ticalanalysispresentedinFig.3.TheaveragePvaluefortherankcorrelationsbetweenccr4-⌬experimentsϪ6.5andtranscrip-tionalshutoffexperimentsisonly10,whichliesinthecentral50%ofpairwiserankcorrelationsamongtheWuetal.referenceset(61)(Fig.2B).Asimilarresultwasobtainedwhenwecorrectedforthealterationsinthefirsttimepoint(i.e.,subtractedtheratioforthefirsttimepointfromallsub-sequenttimepointsforeachgene)orifweexaminedtheranksoftheslopesoftheratiosoverthetimecourse(datanotshown).Thismakesitunlikelythattheeffectsseeninccr4-⌬reflectmerelyauniformincreaseoflifetimes(whichwouldleadtoapparentup-regulationofunstabletranscriptsanddown-regulationofstabletranscriptsinourexperimentalpro-tocolwithequalamountsoftotalmRNAsfromthemutantandthewildtypeforhybridization).

Twoclassesoftranscriptsthatdoappeartobestabilizedbyccr4-⌬correspondtothosethatencoderibosomalproteinsandthosethatencoderibosomebiogenesisfactors(Fig.5B,purpleandbluebars,respectively).Thesetranscriptsareeasilyiden-tifiedbecausetheyarealmostcompletelyunaffectedatthezerotimepointandtheirratiograduallyincreasesovertime(inlatetimepoints,thesignalinoneorbothchannelsislostandmicroarraymeasurementsbecomeunreliable;theeffectismorepronouncedforribosomebiogenesisfactorswhichhavetheshortestmRNAhalf-lives)(Fig.5B).Figure6showstheindividualgenesforthesetwogroups.

Therewerestrikingdifferencesintheclassesoftranscriptsthatwereaffectedinthemutants.Inthepub1-⌬strain,asintheccr4-⌬strain,transcriptsencodingmitochondrialproteinsdisplayedalteredabundanceatthezerotimepoint,andtherewassimilarlylittlechangeinrelativeratiosofthesetranscriptsoverthetimecourse.Incontrasttoccr4-⌬,however,pub1-⌬displayedpreferentialstabilizationoftranscriptsencodingri-bosomalproteins.Inpan2-⌬andpuf4-⌬,therewerefewalter-ationsatthezerotimepoint,whilethestabilizationoftran-scriptsencodingribosomalproteinsandribosomebiogenesisfactorswassimilartothatofccr4-⌬,butlesspenetrant.Amongthefourmutants,thedatafrompan2-⌬showedthestrongestrelationshiptothepoly(A)ϩmRNAdecayprofile(Fig.3and

MOL.CELL.BIOL.

5B),indicatingthatitseffectsaremostrelatedtosimplyex-tendingoverallmRNAhalf-lives,consistentwiththenotionthatPan2pactsconstitutivelyonallmRNAs.ThisfindingalsoraisesthepossibilitythatCcr4p,whichisthemajoryeastdead-enylase(54),hasapreferencefortranscriptsencodingribo-somebiogenesisfactorsandribosomalproteins.

Thesedatashowthatthereisinformationthatcanbeob-tainedonlybyexaminingthechangesinthetranscriptratiosoverthetimecourse,validatingtheexperimentalapproachshowninFig.5A,whichwastheoriginalaimofthisstudy.Noneoftheeffectsontranscriptsencodingnucleolarorribo-somalproteinsareevidentfromcomparisonsofthemutantversusthewildtypeatthezerotimepoint,indicatingthatthetimecoursefollowingtranscriptionalshutoff(outlinedinFig.5A)isnecessarytodetecttheinfluenceofeachofthesefactorsonthedecayofthesetranscripts.Infact,thezerotimepointanalysis(i.e.,steady-statecomparison)maybemisleading;onewouldconcludethatpub1-⌬andccr4-⌬primarilycausealter-ationsintheabundanceofavarietyofmitochondrialtran-scripts(Fig.5B),butthesedonotappeartobeduetoanalterationinstabilityofthesetranscripts.

Ccr4pisamediatoroftheheatshockresponse.SincethetranscriptionalresponsetoheatshockconsistslargelyofatranscriptionalshutofffollowedbymRNAdecay(39),andsinceCcr4pappearstopreferentiallycontributetothedegra-dationoftranscriptswhoseabundanceisknowntodecreaseinresponsetoheatshock(15),wereasonedthatCcr4pislikelytoberequiredforthisaspectoftheheatshockresponse.Totestthis,weexaminedtheeffectsofccr4-⌬onmRNAlevelsfol-lowingatimecourseafterheatshock(Fig.5B).Theresultsweresimilartothoseobtainedbytreatmentoftheccr4-⌬mutantwith1,10-phenanthrolineand6-azauracilastheinhib-itoroftranscription;whilethestabilityofmosttranscriptswaslargelyunaffected,aclearstabilizationoftranscriptsencodingribosomalproteinsandribosomebiogenesisfactorswasob-served.ThisresultshowsthatCcr4pisamediatoroftheheatshockresponse.

DISCUSSION

Evidencethatdiversebiologicalprocessesareregulatedbycontrolledtranscriptdecaycontinuestoaccumulateinthelit-erature.Forexample,nocturnin,whichisrelatedtotheyeastCcr4pdeadenylase,isspecificallyexpressedintheXenopuslaevisretinaandcontrolsmelatoninlevels(5).Incontrasttothehundredsofknownsequence-specifictranscriptionalacti-vators,however,relativelyfewfactorshavebeenidentifiedthatregulatethestabilityofdifferentclassesoffunctionallyrelatedmRNAtranscripts.ThesubstraterangehasnotbeenreportedformostofthefactorsalreadyimplicatedinmRNAstability,eveninyeast,anorganismforwhichobtainingdeletionmu-tantsisnowtrivial(16).

RegulationofnucleolarproteinsatthelevelofmRNAsta-bility.AmongthespecificfunctionalcategoriesthatfromouranalysesappeartoberegulatedatthelevelofmRNAstabilityarerRNAbiogenesisandribosomeassembly.Manyoftheproteinsinthesecategoriesarelocatedinthenucleolus.TherRNAbiogenesisandribosomeassemblycategoriesoverlapwithoneanotherbutaredistinctfromtheribosomalproteins.Althoughyeastribosomebiosynthesisisalreadyknowntobe

Downloaded from mcb.asm.org at Penn State Univ on February 5, 2008 shownFIG.in6.Fig.Individual5B.Phen.,transcripts1,10-phenanthroline;encodingnucleolar6-azaU,proteins6-azauracil.

(top)andribosomalproteins(bottom),correspondingtothepurpleandbluebars5545

Downloaded from mcb.asm.org at Penn State Univ on February 5, 2008 5546GRIGULLETAL.regulatedatmultiplelevels(38,57,59),ithasnottoourknowledgebeenpreviouslyreportedthatyeastnucleolarpro-teinsareregulatedatthelevelofmRNAstability.Infact,theobservationthattheabundanceofthesetranscriptsislargelyunchangedfromthatofthewildtypeatthezerotimepoint(Fig.5and6)suggeststhattheirtranscriptionratemaybedown-regulatedinthesemutantstocompensatefortheirin-creasedstability.Ourdata,andalsothoseofWangetal.(58)(Fig.2B)indicatethattheseareamongtheleaststabletran-scriptsinthecell,atleastupontheinhibitionoftranscription.Wehaveidentifiedthreepotentialtrans-actingfactorsthataffecttherelativedegradationratesofthesetranscripts:Ccr4p,Pan2p,andPuf4p,whichallappeartocontributetotheirdegradation.Pub1p,however,appearstobeastabilizingfactorforribosomalproteintranscripts,whichhavepreviouslybeennotedtohaveshorthalf-livesupontranscriptionalshutoff(20,58).

Rapiddown-regulationoftranscriptsencodingnucleolarproteins(andalsoribosomalproteins)isafeatureofmanymicroarrayanalysesofvariousperturbationsofyeastintheliterature(15,61).Transcriptionalcoregulationofthesegeneshaspreviouslybeenattributedtotwodifferentpromoterele-ments(PACandRRPE)sharedbymanyofthesegenes(50).Theinherentinstabilityofthesetranscriptswouldexplaintheirrapiddeclineinabundanceupondown-regulationoftranscrip-tion.However,itisalsopossiblethattheirstabilityisitselfpartoftheregulatorymechanism;thispossibilityissupportedbythefactthatthreeofthemutantsweanalyzedaffectedthedeadenylationrateofthesetranscriptsandalsothattheirdeadenylationappearsmorerapidinresponsetoheatshockthantotranscriptioninhibition,onthebasisoftheccr4-⌬datainFig.5B.Regardless,themRNAdegradationrateplaysanimportantroleintheregulationofthisgroupoftranscripts,asweshowbyanalysisoftheheatshockresponseinthepresenceandabsenceofccr4-⌬.Hence,thisgroupoftranscriptsappearstorepresent,inatleastonesense,theposttranscriptionalfunc-tionalequivalentofabacterialoperon,asproposedbyKeeneandTenenbaum(30,31).

DoestranscriptionalshutoffitselfimpactmRNAdecay?Sincetheseexperimentswereconductedoveratimecourseofinhibitionoftranscription,onecaveatofourresultsisthattheymayberelevantonlytothiscondition(e.g.,theymaycorre-spondtoposttranscriptionalmechanismsthatreacttothelossofgeneraltranscription).Thesameargumentappliestomanystudiesintheliteraturethathaveutilizedinhibitionoftran-scriptiontomonitormRNAdecay(20,38,41,58).ApreviouscomparisonofmRNAhalf-livesmeasuredbybothinvivoradiolabelingandNorthernblottingfollowingtranscriptionalshutoffsuggestedthatthereisaloosebutpositivecorrelationbetweenhalf-livesmeasuredwithandwithoutshutoff(20).However,onlyasmallnumberofgenes(11,plus3unnamedcDNAs)werecompared.Itmayultimatelybeworthwhiletoperformsuchexperimentsonagenome-widescale.

Anonlinetoolforidentifyingbiologicallysignificanttrendsinrelativeranksdata.TheKStestprovidesagenerallyappli-cablemethodforidentifyingcategoriesoftranscriptsthataresignificantlystabilizedordestabilizedunderanygivencondi-tion.Fortheconvenienceofresearcherswishingtoconfirmourfindingsoranalyzetheirowndata,wehaveimplementedaweb-accessibleversionoftheKStestathttp://kstest.med

MOL.CELL.BIOL.

.utoronto.ca,inwhichlistsofrankedyeastgenenamescanbeenteredandasummaryofKStestPvaluescanberetrieved.TheKStestisusefulforanalyzingyeastdataofanytypethatcanbesorted,includingnotonlymicroarrayratiosbutothertypesofgenomicdatasuchasabsolutelevelsofgeneorproteinexpression,colonysizes,orevenscreeningresults.

IsthestabilityofyeastnucleolarproteinsregulatedbyUTRelements?Avarietyofcomputationaltoolscanalsobeappliedtoidentifypossiblecis-actingelementsinUTRsoftranscriptswithsharedstabilitycharacteristics(3,23,28,35).Thepoten-tialofsuchanalyseshasrecentlybeendemonstratedinArabi-dopsisthaliana(6).However,wehaveasyetbeenunabletoidentifyanystatisticallysignificantelementssharedbyama-jorityofthe3Јor5ЈUTRsamongthedifferentgroupsoffunctionallyrelatedtranscriptsdescribedhere(Fig.1B,4,and5B).

Globalanalysisofpotentialtrans-actingmRNAstabilityfactors.Themethodsdescribedhere(inparticular,theproce-dureshowninFig.5A)canbeusedtodeterminewhetheramutationinanyproteinknownorsuspectedtobeinvolvedinmRNAstabilitypreferentiallystabilizesordestabilizescertainclassesoftranscripts.Despiteitseffectinup-regulatingasmallgroupoftranscriptsenrichedinfunctionsrelatedtoionho-meostasis(Fig.4),1,10-phenanthrolineisthemostpracticalofthefivechemicalswetestedforapplyingthismethodinyeast.Itisinexpensiveandeasilyobtained,themechanismofactionisunderstood,anditappearstoworkslightlyfasterthantheotherchemicalswetested(Fig.1B).Thiolutinhasaverysim-ilareffect(Fig.1B,3,and4)andiseffectiveatlowerconcen-trations,butitisnotavailablecommercially.Toourknowl-edge,theprecisemodeofactionofthiolutinisunknown,althoughourresults(Fig.4)suggestthatitmayactinthesamewayas1,10-phenanthroline.

IncontrasttomicroarrayanalysisofmRNApopulationsbetweenthemutantandthewildtypeatsteadystate,ourapproachcapturesagreaternumberofdifferentialtranscripts(Fig.5B)andaffordshigherconfidenceandstatisticalpowerthroughtheidentificationoftranscriptswhoserelativeabun-dancechangesconsistentlyacrossthetimecourse.Incompar-isontomeasuringthepoly(A)ϩdecaytimecoursesofwild-typeandmutantculturesindependently,thismethodalsotakesadvantageoftheabilityofthetwo-colorsystemtohighlightrelativelysmalldifferencesinrelativetranscriptabundance.ItwillbeofinteresttoanalyzethemultitudeoffactorsknownorsuspectedtoaffectstabilityofmRNAs.

ACKNOWLEDGMENTS

andThistheworkConnaughtwassupportedFoundationbygrants(UniversityfromGenomeofToronto)Canada,toNSERC,evaluationWethankforofRickthemanuscript,Collins,JimBenIngles,BlencoweandAlanT.R.H.forCochraneforcriticalYangthegiftofthiolutin,DaveJansmafortherpb1-1helpfulstrain,advice,andStuartPfizerport.

andNaveedMohammadfortechnicalandcomputationalsup-REFERENCES

1.Adams,C.C.,andD.S.Gross.1991.Theyeastheatshockresponseisinducedbyconversionofcellstospheroplastsandbypotenttranscriptionalinhibitors.J.Bacteriol.173:7429–7435.

2.Allison,L.A.,J.K.Wong,V.D.Fitzpatrick,M.Moyle,andC.J.Ingles.1988.TheSaccharomycesC-terminalcerevisiaedomain,ofDrosophilathelargestmelanogastersubunitof,RNAandmammals:polymeraseaIIcon-ofservedstructurewithanessentialfunction.Mol.Cell.Biol.8:321–329.

Downloaded from mcb.asm.org at Penn State Univ on February 5, 2008 VOL.24,2004

3.Anderson,cis-actingelementsJ.S.J.,affectingandR.Parker.post-transcriptional2000.ComputationalcontrolofidentigenefiexpressioncationofinSaccharomycescerevisiae.NucleicAcidsRes.28:1604–1617.

4.Ashburner,M.,C.A.Ball,J.A.Blake,D.Botstein,H.Butler,etal.2000.Geneontology:toolfortheunificationofbiology.Nat.Genet.25:25–29.5.Baggs,laevisretina:J.E.,andamechanismC.B.Green.for2003.posttranscriptionalNocturnin,adeadenylasecontrolofcircadian-re-inXenopuslatedmRNA.Curr.Biol.13:1–198.

6.Bateman,A.2002.TheSGS3proteininvolvedinPTGSfindsafamily.BMCBioinformatics3:21.

7.Belasco,J.G.,andG.Brawerman(ed.).1993.ControlofmessengerRNAstability.AcademicPress,SanDiego,Calif.

8.Bernstein,J.A.,A.B.Khodursky,P.H.Lin,S.Lin-Chao,andS.N.Cohen.2001.GlobalanalysisofmRNAdecayandabundanceinEscherichiacoliatsingle-generesolutionusingtwo-colorfluorescentDNAmicroarrays.Proc.Natl.Acad.Sci.USA99:9697–9702.

9.Blaxall,B.C.,A.Pende,S.C.Wu,andJ.D.Port.2002.CorrelationbetweenintrinsicmRNAstabilityandtheaffinityofAUF1(hnRNPD)andHuRforAϩU-richmRNAs.Mol.Cell.Biochem.232:1–11.

10.Cao,D.,andR.Parker.2001.ComputationalmodelingofeukaryoticmRNA

turnover.RNA7:1192–1212.

11.Causton,H.C.,B.Ren,S.S.Koh,P.T.Harbison,E.Kanin,E.G.Jennings,

T.I.Lee,H.L.True,E.S.Lander,andR.A.Young.2001.Remodelingofyeastgenomeexpressioninresponsetoenvironmentalchanges.Mol.Biol.Cell12:323–337.

12.Chen,C.Y.,andA.B.Shyu.1995.AU-richelements:characterizationand

importanceinmRNAdegradation.TrendsBiochem.Sci.20:465–470.

13.Chen,C.Y.,R.Gherzi,S.E.Ong,E.L.Chan,R.Rajmakers,G.J.Prujin,

G.Stoecklin,C.Moroni,M.Mann,andM.Karin.2001.AUbindingproteinsrecruittheexosometodegradeARE-containingmRNAs.Cell107:451–4.14.Exinger,F.,andF.Lacroute.1992.6-AzauracilinhibitionofGTPbiosyn-thesisinSaccharomycescerevisiae.Curr.Genet.22:9–11.

15.Gasch,A.P.,P.T.Spellman,C.M.Kao,O.Carmel-Harel,M.B.Eisen,G.

Storz,D.Botstein,andP.O.Brown.2000.Genomicexpressionprogramsintheresponseofyeastcellstoenvironmentalchanges.Mol.Biol.Cell11:4241–4257.

16.Giaever,G.,A.M.Chu,L.Ni,C.Connelly,L.Riles,etal.2002.Functional

profilingoftheSaccharomycescerevisiaegenome.Nature418:387–391.17.Gibbons,J.D.,andS.Chakraborti.1992.Nonparametricstatisticalinfer-ence,3rded.MarcelDekkerInc.,NewYork,N.Y.

18.Gutierrez,R.A.,R.M.Ewing,J.M.Cherry,andP.J.Green.2002.Identi-ficationofunstabletranscriptsinArabidopsisbycDNAmicroarrayanalysis:rapiddecayisassociatedwithagroupoftouch-andspecificclock-controlledgenes.Proc.Natl.Acad.Sci.USA99:11513–11518.

19.Hegde,P.,R.Qi,K.Abernathy,C.Gay,S.Dharap,etal.2000.Aconcise

guidetocDNAmicroarrayanalysis.BioTechniques29:548–556.

20.Herrick,D.,R.Parker,andA.Jacobson1990.Identificationandcomparison

ofstableandunstablemRNAsinSaccharomycescerevisiae.Mol.Cell.Biol.10:2269–2284.

21.Herruer,M.H.,W.H.Mager,H.A.Raue,P.Vreken,E.Wilms,andR.J.

Planta.1988.MildtemperatureshockaffectstranscriptionofyeastribosomalproteingenesaswellasthestabilityoftheirmRNAs.NucleicAcidsRes.16:7917–7929.

22.Hsu,C.L.,andA.Stevens.1993.Yeastcellslacking5Ј33Јexoribonuclease

1containmRNAspeciesthatarepoly(A)deficientandpartiallylackthe5Јcapstructure.Mol.Cell.Biol.13:4826–4835.

23.Hu,Y.J.2002.Predictionofstructuralmotifsinafamilyofcoregulated

RNAsequences.NucleicAcidsRes.30:3886–33.

24.Hughes,T.R.,M.J.Marton,A.R.Jones,C.J.Roberts,R.Stoughton,C.D.

Armour,H.A.Bennett,E.Coffey,H.Dai,Y.D.He,M.J.Kidd,A.M.King,M.R.Meyer,D.Slade,P.Y.Lum,S.B.Stepaniants,D.D.Shoemaker,D.Gachotte,K.Chakraburtty,J.Simon,M.Bard,andS.H.Friend.2000.Func-tionaldiscoveryviaacompendiumofexpressionprofiles.Cell102:109–126.25.Issel-Tarver,L.,K.R.Christie,K.Dolinski,R.Andrada,R.Balakrishnan,

C.A.Ball,G.Binkley,S.Dong,S.S.Dwight,D.G.Fisk,M.Harris,M.Schroeder,A.Sethuraman,K.Tse,S.Weng,D.Botstein,andJ.M.Cherry.2002.SaccharomycesGenomeDatabase.MethodsEnzymol.350:329–346.26.Jacobs,J.S.,A.R.Anderson,andR.P.Parker.1998.The3Јto5Јdegra-dationofyeastmRNAsisageneralmechanismformRNAturnoverthatrequirestheSKI2DEVHboxproteinand3Јto5Јexonucleasesoftheexosomecomplex.EMBOJ.17:1497–1506.

27.Jacobson,A.,andS.W.Peltz.1996.Interrelationshipsofthepathwaysof

mRNAdecayandtranslationineukaryoticcells.Annu.Rev.Biochem.65:693–773.

28.Jacobson,A.B.,andM.Zuker.1993.Structuralanalysisbyenergydotplot

ofalargemRNA.J.Mol.Biol.233:261–269.

29.Johnston,J.R.1994.Moleculargeneticsofyeast:apracticalapproach.

OxfordUniversityPress,NewYork,N.Y.

30.Keene,J.D.2001.Ribonucleoproteininfrastructureregulatingtheflowof

geneticinformationbetweenthegenomeandtheproteome.Proc.Natl.Acad.Sci.USA98:7018–7024.

TRANSCRIPTIONINHIBITORSANDMICROARRAYS5547

31.Keene,J.D.,andS.A.Tenenbaum.2002.EukaryoticmRNPsmayrepresent

posttranscriptionaloperons.Mol.Cell9:1161–1167.

32.Kief,D.R.,andJ.R.Warner.1981.Hierarchyofelementsregulatingsyn-thesisofribosomalproteinsinSaccharomycescerevisiae.Mol.Cell.Biol.1:1016–1023.

33.LaGrandeur,T.E.,andR.Parker.1998.Isolationandcharacterizationof

Dcp1,theyeastmRNAdecappingenzyme.EMBOJ.17:1487–1496.

34.LaGrandeur,T.E.,andR.Parker.1999.Thecisactingsequencesrespon-sibleforthedifferentialdecayoftheunstableMFA2andstablePGK1transcriptsinyeastincludethecontextorthetranslationalstartcodon.RNA5:420–433.

35.Le,S.Y.,J.H.Chen,D.Konings,andJ.V.Maizel,Jr.2003.Discovering

well-orderedfoldingpatternsinnucleotidesequences.Bioinformatics19:354–361.

36.Lehmann,E.L.,andH.J.M.D’Abrera.1998.Nonparametrics:statistical

methodsbasedonranks,reviseded.Prentice-Hall,EnglewoodCliffs,N.J.37.Lelivelt,M.J.,andM.R.Culbertson.1999.YeastUpfproteinsforRNA

surveillanceaffectglobalexpressionoftheyeasttranscriptome.Mol.Cell.Biol.19:6710–6719.

38.Li,B.,C.R.Nierras,andJ.R.Warner.1999.Transcriptionalelements

involvedintherepressionofribosomalproteinsynthesis.Mol.Cell.Biol.19:5393–5404.

39.Lindquist,S.1981.Regulationofproteinsynthesisduringheatshock.Na-ture293:311–314.

40.Mignone,F.,C.Gissi,S.Liuni,andG.Pesole.2002.Untranslatedregionsof

mRNAs.GenomeBiol.3:REVIEWS0004.1–REVIEWS0004.10.[Online.]http://genomebiology.com/2002/3/3/REVIEWS/0004.

41.Moore,P.A.,F.A.Sagliocco,R.M.Wood,andA.J.Brown.1991.Yeast

glycolyticmRNAsaredifferentiallyregulated.Mol.Cell.Biol.11:5330–5337.42.Nonet,M.,C.Scafe,J.Sexton,andR.Young.1987.EukaryoticRNApoly-meraseconditionalmutantthatrapidlyceasesmRNAsynthesis.Mol.Cell.Biol.7:1602–1611.

43.Olivas,W.,andR.Parker.2000.ThePuf3proteinisatranscript-specific

regulatorofmRNAdegradationinyeast.EMBOJ.19:6602–6611.

44.Parker,R.,andH.Song.2004.Theenzymesandcontrolofeukaryotic

mRNAturnover.Nat.Struct.Mol.Biol.11:121–127.

45.Raghavan,A.,R.L.Ogilvie,C.Reilly,M.L.Abelson,S.Raghavan,J.

Vasdevani,M.Krathwol,andP.R.Bohjanen.2002.Genome-wideanalysisofmRNAdecayinrestingandactivatedprimaryhumanTlymphocytes.NucleicAcidsRes.30:5529–5538.

46.Robinson,M.D.,J.Grigull,N.Mohammad,andT.R.Hughes.2002.Fun-Spec:aweb-basedclusterinterpreterforyeast.BMCBioinformatics3:35.47.Sachs,L.1982.Appliedstatistics.Springer-Verlag,NewYork,N.Y.

48.Shaw,R.J.,andD.Reines.Saccharomycescerevisiaetranscriptionelongation

mutantsaredefectiveinPUR5inductioninresponsetonucleotidedeple-tion.Mol.Cell.Biol.20:7427–7437.

49.Straney,D.C.,andD.M.Crothers.1987.Effectofdrug-DNAinteractions

upontranscriptioninitiationatthelacpromoter.Biochemistry26:1987–1995.50.Tavazoie,S.,J.D.Hughes,M.J.Campbell,R.J.Cho,andG.M.Church.

1999.Systematicdeterminationofgeneticnetworkarchitecture.Nat.Genet.22:281–285.

51.Theil,E.C.1993.TheIRE(ironregulatoryelement)family:structureswhich

regulatemRNAtranslationorstability.Biofactors4:87–93.

52.Tipper,D.J.1973.Inhibitionofyeastribonucleicacidpolymerasesbythio-lutin.J.Bacteriol.116:245–256.

53.Tucker,M.,andR.Parker.2000.MechanismsandcontrolofmRNAdecap-pinginSaccharomycescerevisiae.Annu.Rev.Biochem.69:571–595.

54.Tucker,M.,M.A.Valencia-Sanchez,R.R.Staples,J.Chen,C.L.Denis,and

R.Parker.2001.ThetranscriptionfactorassociatedCcr4andCaf1proteinsaremycescomponentscerevisiae.ofCellthe104:major377–cytoplasmic386.

mRNAdeadenylaseinSaccharo-55.Tucker,M.,R.R.Staples,M.A.Valencia-Sanchez,D.Muhlrad,andR.

Parker.2002.Ccr4pisthecatalyticsubunitofaCcr4p/Pop2p/NotpmRNAdeadenylasecomplexinSaccharomycescerevisiae.EMBOJ.21:1427–1436.56.Vasudevan,S.,andS.W.Peltz.2001.RegulatedARE-mediatedmRNA

decayinSaccharomycescerevisiae.Mol.Cell7:1191–1200.

57.Venema,cerevisiae.J.,Annu.andD.Rev.Tollervey.Genet.33:1999.261–Ribosome311.

synthesisinSaccharomyces

58.Wang,Y.,C.L.Liu,J.D.Storey,R.J.Tibshirani,D.Herschlag,andP.O.

Brown.2002.PrecisionandfunctionalspecificityinmRNAdecay.Proc.Natl.Acad.Sci.USA99:5860–5865.

59.Warner,J.R.1999.Theeconomicsofribosomebiosynthesisinyeast.Trends

Biochem.Sci.24:437–440.

60.Wilusz,C.J.,M.Wormington,andS.W.Peltz.2001.Thecap-to-tailguide

tomRNAturnover.Nat.Rev.Mol.Cell.Biol.2:237–246.

61.Wu,L.F.,T.R.Hughes,A.P.Davierwala,M.D.Robinson,R.Stoughton,and

S.J.Altschuler.2002.Large-scalepredictionofSaccharomycescerevisiaegenefunctionusingoverlappingtranscriptionalclusters.Nat.Genet.31:255–265.62.Yang,Y.H.,S.Dudoit,P.Luu,D.M.Lin,V.Peng,J.Ngai,andT.P.Speed.

2002.NormalizationforcDNAmicroarraydata:arobustcompositemethodaddressingsingleandmultipleslidesystematicvariation.NucleicAcidsRes.30:e15.

Downloaded from mcb.asm.org at Penn State Univ on February 5, 2008